Comments (16)
@winterwang I think this is a different issue from #45. Is there any way you would be able to share the data you are using? It does not even need to be the full data---it can be a single chromosome, for example. I'd like to run some tests to understand what is going on.
from gemma.
@pcarbo Thanks for the reply
from gemma.
@winterwang I was able to successfully run this command on my MacBook Pro:
~/git/gemma/bin/gemma -g GEMMAchr22txt.gz -p GEMMApheno.txt \
-n 1 2 3 4 -k GEMMA_standardrelatedmat005.sXX.txt -lmm
See the attached log file.
What OS are you using? What version of gemma are you using?
from gemma.
@pcarbo Thanks for the feedback.
However, I still get nan for the command.
Previously I used latest version 0.97 downloaded from github.com, compiled and tested on my Ubuntu 16.04 machine:
*********************************************************
Genome-wide Efficient Mixed Model Association (GEMMA)
Version 0.97, 07/07/2017
Visit http://www.xzlab.org/software.html For Updates
(C) 2017 Xiang Zhou
GNU General Public License
For Help, Type ./gemma -h
*********************************************************
I found you are using version 0.96 on a Mac Pro. So I switched to version 0.96, the problem still not fixed still on my Ubuntu 16.04 platform:
*********************************************************
Genome-wide Efficient Mixed Model Association (GEMMA)
Version 0.96, 05/17/2017
Visit http://www.xzlab.org/software.html For Updates
(C) 2017 Xiang Zhou
GNU General Public License
For Help, Type ./gemma -h
*********************************************************
./gemma -g/GEMMAchr22txt.gz -p GEMMApheno.txt -n 1 2 3 4 -k GEMMA_centerrelatedmat005.cXX.txt -lmm
Reading Files ...
## number of total individuals = 807
## number of analyzed individuals = 795
## number of covariates = 1
## number of phenotypes = 4
## number of total SNPs = 163763
## number of analyzed SNPs = 114095
Start Eigen-Decomposition...
REMLE estimate for Vg in the null model:
-nan
-nan -nan
-nan -nan -nan
-nan -nan -nan -nan
se(Vg):
-nan
-nan -nan
-nan -nan -nan
-nan -nan -nan -nan
REMLE estimate for Ve in the null model:
-nan
-nan -nan
-nan -nan -nan
-nan -nan -nan -nan
se(Ve):
-nan
-nan -nan
-nan -nan -nan
-nan -nan -nan -nan
REMLE likelihood = -nan
MLE estimate for Vg in the null model:
-nan
-nan -nan
-nan -nan -nan
-nan -nan -nan -nan
se(Vg):
-nan
-nan -nan
-nan -nan -nan
-nan -nan -nan -nan
MLE estimate for Ve in the null model:
-nan
-nan -nan
-nan -nan -nan
-nan -nan -nan -nan
se(Ve):
-nan
-nan -nan
-nan -nan -nan
-nan -nan -nan -nan
MLE likelihood = -nan
Reading SNPs 0.00%
Reading SNPs kept 0.00% and not moving forward.
Can you succeed when using all of the 5 phenotypes on your Mac Pro?
from gemma.
@winterwang I was able to reproduce the NaN's using gemma v0.96 for Linux. It is worrying that we have a platform-dependent issue! I will look in to it.
from gemma.
@winterwang The discovered the problem: the Linux binary provided in the GEMMA 0.96 release was linked to GSL 2.2.1, and it seems that it only works with GSL 1.x, so I linked with GSL 1.16 instead. Now I am no longer getting the NaN's. I've attached the new binary, and we will soon make a GEMMA 0.97 release available for download.
Can you please test that the attached binary works for you on your Ubuntu system?
from gemma.
Also, I updated the binary for the v0.96 release.
from gemma.
Hi, @pcarbo . Your help is greatly appreciated. Now using the new binary you attached for me I am able to compute the command with 4 and 5 phenotypes on my Ubuntu 16.04. No more nan is created.
Thank you so much!
##
## GEMMA Version = 0.97
##
## Command Line Input = ./gemma.linux -g GEMMAchr22txt.gz -p GEMMApheno.txt -n 1 2 3 4 -k ./output/GEMMA_centerrelatedmat005.cXX.txt -lmm -o test
##
## Date = Fri Jul 28 09:14:00 2017
##
## Summary Statistics:
## number of total individuals = 807
## number of analyzed individuals = 795
## number of covariates = 1
## number of phenotypes = 4
## number of total SNPs = 163763
## number of analyzed SNPs = 114095
## REMLE log-likelihood in the null model = -4299.15
## MLE log-likelihood in the null model = -4302.56
## REMLE estimate for Vg in the null model:
9.96209e-06
7.58463e-14 9.71815e-06
-7.87707e-13 1.66e-11 9.56255e-06
4.365e-13 1.44716e-12 -1.9888e-13 9.37882e-06
## se(Vg):
1.06002
0.760842 1.08813
0.722019 0.88268 0.981513
0.663478 0.669846 0.640012 0.841037
## REMLE estimate for Ve in the null model:
0.996209
0.00285371 0.971815
-0.01166 0.538287 0.956255
0.0264271 0.0530889 -0.0166509 0.937882
## se(Ve):
0.333927
0.237622 0.337159
0.227644 0.27476 0.309713
0.209418 0.209679 0.202253 0.266001
## MLE estimate for Vg in the null model:
9.96208e-06 1.26323e-13 -1.08012e-12 1.76122e-13
1.26323e-13 9.71815e-06 1.63898e-11 1.25851e-12
-1.08012e-12 1.63898e-11 9.56255e-06 7.06729e-14
1.76122e-13 1.25851e-12 7.06729e-14 9.37882e-06
## se(Vg):
1.08775
0.781956 1.11965
0.74072 0.908896 1.00651
0.67756 0.684627 0.653495 0.85589
## MLE estimate for Ve in the null model:
0.994956 0.00285012 -0.0116453 0.0263939
0.00285012 0.970592 0.53761 0.0530221
-0.0116453 0.53761 0.955052 -0.01663
0.0263939 0.0530221 -0.01663 0.936702
## se(Ve):
0.342888
0.244368 0.347138
0.233676 0.28309 0.317762
0.214008 0.214445 0.20664 0.270878
## estimate for B (d by c) in the null model (columns correspond to the covariates provided in the file):
-0.00670195
-0.00196422
0.00190655
-0.00429808
## se(B):
0.035399
0.034963
0.0346819
0.0343471
##
## Computation Time:
## total computation time = 7.03936 min
## computation time break down:
## time on eigen-decomposition = 0.00856403 min
## time on calculating UtX = 0.162786 min
## time on optimization = 6.42819 min
##
from gemma.
Can we add a test to the test suite?
from gemma.
@pjotrp Good idea. I will do that.
from gemma.
@pjotrp I'm having trouble reproducing the error we had earlier, and I no longer have the GEMMA binary linked to the GSL 2.x library. I suggest we close this issue---I don't think this requires an additional test.
from gemma.
I can test GSL versions with GNU Guix. We had the same problem in GeneNetwork.org - I'll try to find a dataset.
from gemma.
I have reproduced the problem. Will add test data to GEMMA soon - is a plink file which is good too.
from gemma.
@pjotrp Great!
from gemma.
Looks like the fix of issue 26 also fixes the GSLv2 for this issue. Test that fails on master coming up.
from gemma.
This has been fixed and tests added in #92. Note that if you get NaNs again, it is probably due to highly correlated phenotypes/covariates. Please reopen if an issue resurfaces.
from gemma.
Related Issues (20)
- Enforce failed for number of analyzed individuals equals 0. in src/param.cpp at line 2073 in ProcessCvtPhen HOT 2
- long haplotype as population structure
- basic usage doesn't work
- Gemma run-time for Estimate Relatedness Matrix calculation
- very low number of analyzed SNPs/var
- The input files includes 26 linkage groups but the output files only includes 22 linkage group
- Categorical covariates
- Pve is .999 HOT 1
- gemma-wrapper:279:in `<main>': Expected GEMMA -a genotype file switch (RuntimeError)
- Got a nan in the p-value results HOT 1
- gsl: newton.c:88: ERROR: function value is not finite 3.08%
- GSL ERROR: function value is not finite in brent.c at line 58 errno 9
- WARNING: Brent did not converge
- Why the diagonal of relatedness matrix is not equal to 1 or a constant value? HOT 1
- The problem of filtering snp parameters
- ERROR: Enforce failed for Problem reading FAM file (phenotypes do not match geno file) in src/gemma_io.cpp at line 600 in ReadFile_fam GEMMA 0.98.3 (2020-11-28) by Xiang Zhou and team (C) 2012-2020
- Identical P-Values at end of run HOT 1
- BIMBAM format (preferred)
- ##ERROR: FAILED: Parsing input file 'mikk_small_hapVariants_updated_noheaders.csv' failed in function ReadFile_geno in src/gemma_io.cpp at line 744, for GEMMA 0.98.5 (2021-08-25) by Xiang Zhou, Pjotr Prins and team (C) 2012-2021
- GEMMA for UK Biobank genome-wide analysis?
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from gemma.